There are problems of noise interference in low-frequency oscillation pattern recognition and parameter extraction in a power system. Thus a new method for extracting key modal parameters of low-frequency oscillation is proposed, one which combines a tunable Q factor wavelet transform (TQWT) with a sparse time domain (STD) method. First, TQWT technology is used to preprocess the wide-area measurement signal of low-frequency oscillation in a power system with noise, and then the processed signal is used as a new input signal to identify the oscillation modes and their parameters by an STD algorithm. Then the input signal can be collected by single-point or multi-point measurement. The advantages of the proposed method are verified by simulation of the test signal and a EPRI-36 machine system. It can effectively suppress the noise and accurately identify the oscillation modal parameters of the system in the environment with low signal-to-noise ratio. Compared with the traditional method, it has better anti-noise ability, shorter identification time and more accurate identified parameters. This work is supported by the National Natural Science Foundation of China (No. 51977039).